Use of Selective Cyclooxygenase-2 Inhibitors, Other Analgesics, and Risk of Glioma

(1)

Use of Selective Cyclooxygenase-2 Inhibitors, Other Analgesics, and Risk of Glioma

Corinna Seliger¹*, Christoph R. Meier^2,3,4, Claudia Becker², Susan S. Jick³, Ulrich Bogdahn¹, Peter Hau¹, Michael F. Leitzmann⁵

1Department of Neurology and Wilhelm Sander-NeuroOncology Unit, Regensburg University Hospital, Regensburg, Germany,2Basel Pharmacoepidemiology Unit, Division of CIinical Pharmacy and Epidemiology, Department of Pharmaceutical Sciences, University of Basel, Basel, Switzerland,3Boston Collaborative Drug Surveillance Program, Boston University School of Public Health, Boston,

Massachusetts, United States of America,4Hospital Pharmacy, University Hospital Basel, Basel, Switzerland,5Department of Epidemiology and Preventive Medicine, University of Regensburg, Regensburg, Germany

*Corinna.Seliger@ukr.de

Abstract

Background

Selective cyclooxygenase-2 (COX-2) inhibitors are analgesic, antipyretic, and anti-inflam- matory drugs. They have been found to inhibit the development of glioma in laboratory investigations. Whether these drugs reduce the risk of glioma incidence in humans is unknown.

Methods

We conducted a matched case-control analysis using the U.K.-based Clinical Practice Research Datalink (CPRD). We identified 2,469 cases matched to 24,690 controls on age, sex, calendar time, general practice, and number of years of active history in the CPRD prior to the index date. We conducted conditional logistic regression analyses to determine relative risks, estimated as odds ratios (ORs) with 95% confidence intervals (CIs) of glioma in relation to use of selective COX-2 inhibitors, adjusted for several confounding variables.

Results

Use of selective COX-2 inhibitors was unrelated to risk of glioma (adjusted OR for 1 – 9 ver- sus 0 prescriptions = 1.02; 95% CI = 0.92 – 1.13, 10 – 29 versus 0 prescriptions = 1.01; 95%

CI = 0.80 – 1.28, 30 versus 0 prescriptions = 1.16; 95% CI = 0.86 – 1.55). Trends for increas- ing numbers of prescriptions for other non-steroidal anti-inflammatory drugs (NSAIDs), and non-NSAID analgesics were also not associated with glioma risk.

Conclusion

Further epidemiologic studies are needed to confirm the null relation of use of selective COX-2 inhibitors to glioma risk and to explain the discrepancy between laboratory investiga- tions and our observational study. Impact: Use of selective COX-2 inhibitors is unrelated to glioma risk.

OPEN ACCESS

Citation:Seliger C, Meier CR, Becker C, Jick SS, Bogdahn U, Hau P, et al. (2016) Use of Selective Cyclooxygenase-2 Inhibitors, Other Analgesics, and Risk of Glioma. PLoS ONE 11(2): e0149293.

doi:10.1371/journal.pone.0149293

Editor:John Wallace, University of Calgary, CANADA

Received:December 15, 2015 Accepted:January 29, 2016 Published:February 12, 2016

Copyright:© 2016 Seliger et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability Statement:All relevant data are within the paper.

Funding:This study was sponsored by the German Research Foundation [KFO 262/P10 to C.S and M.

L.]. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing Interests:The authors have declared that no competing interests exist.

(2)

Introduction

Malignant gliomas are highly aggressive tumours of the central nervous system [1]. Glioblas- toma is the most common and malignant type of glioma and it is associated with a median overall survival of 15 months despite aggressive therapy [1]. In contrast to many other cancers, there are only a few established risk factors for gliomas including increasing age, male gender, Caucasian ethnicity, rare genetic syndromes, and a high-level of ionizing radiation [2].

Selective COX-2 inhibitors are commonly used analgesics frequently prescribed for rheuma- toid arthritis, dysmenorrhoea, or acute pain. Besides their analgesic, anti-inflammatory, and antipyretic effects, they have been found to inhibit or kill glioma cells [3–9], increase radio-sen- sitivity [10 – 15], reduce angiogenesis [14, 16 – 19], and stimulate anti-tumour immune reactions in vitro and in established animal models [16, 20–23]. Clinical trials have tested selective COX- 2 inhibitors as adjuvant therapy for glioma, but results have been inconclusive thus far [17, 24 – 32].

Although selective COX-2 inhibitors reduce gliomagenesis in vitro and in vivo [33], the association between these agents and the risk of glioma has not yet been investigated in obser- vational studies. Prior studies on NSAIDS and glioma, including one study using CPRD data [34], reported null [34–38], inverse [39–42], or positive [43, 44] associations, but those studies did not examine the effect of selective COX-2 inhibitors specifically. The lack of studies on the effect of selective COX-2 inhibitors as a distinct group and the promising biological evidence therefore prompted us to analyse the relation of selective COX-2 inhibitors use to the risk of glioma.

Patients and Methods Data source

The Clinical Practice Research Datalink (CPRD) is a primary care database that holds patient information from around 600 general practices representative of the U.K. population with respect to age, sex, and geographic distribution. Standard coding systems are used by general practitioners to record data on patient diagnoses, prescriptions, hospital admissions, and refer- rals to specialists. Data from office computers are transferred to electronic patient records, as has been described previously [45]. The reliability of diagnostic coding in the CPRD database has been thoroughly validated [46, 47].

The current study was approved by the Independent Scientific Advisory Committee of the CPRD (protocol-number: 15_170). Patient records/information was anonymized and de-iden- tified prior to analysis.

Study population

Case definition. We identified patients with newly diagnosed glioma between 1995 and 2015 using READ codes that are based on the WHO classification of glioma [48] as previously described [49]. To validate glioma diagnoses, we used supporting codes such as codes for pro- cedures, radio- or chemotherapy, and referrals to oncology clinics. Only patients younger than 90 years of age were included in the study population. The ‘index date’ was defined as the date of the first glioma diagnosis minus one year. We shifted the date of diagnosis back by one year for cases and controls to account for the lag time between disease development and detection/

diagnosis, to account for possible changes in medication use prior to the detection and diagno-

sis date of glioma, to control for changes in analgesic treatment, and to account for potential

earlier detection of pre-existing concomitant diseases in case patients caused by early symp-

toms of undiagnosed glioma. In order to increase the likelihood of capturing incident cases

(3)

only and to ensure a sufficient history of medication exposure, we required all patients to have an active history in the database for at least three years prior to the index date. We excluded patients with a history of other concurrent cancers except non-melanoma skin cancer as well as patients with recorded alcoholism or human immunodeficiency virus infection prior to the index date.

Control definition. For each case, we identified at random up to 10 controls without a his- tory of glioma from the CPRD base population matched on calendar time (same index date), age (same year of birth +/- 3 years), sex, general practice, and number of years of active history in the database prior to the index date. We assigned for each control the index date of the cor- responding matching case. To minimize the risk of using control patients with a possible unre- corded glioma diagnosis, we excluded control patients with a prior history of craniotomy within the last year before the index date. Using a one-year time window was considered suffi- cient due to the highly invasive phenotype of malignant glioma. We applied the same exclusion criteria to controls as to cases.

Exposures

We assessed the use of selective COX-2 inhibitors (including etodolac (200 – 1000 mg/day), meloxicam (7.5–15 mg/day), celecoxib (200–400 mg/day), rofecoxib (12.5–25 mg/day), etori- coxib (30 – 120 mg/day), valdecoxib (10 – 40 mg/day), lumiracoxib (100 – 400 mg/day) and diclo- fenac (50–150 mg/day)) from the computerized records. COX-2 selectivity can be determined by dividing 50% inhibitory concentrations (IC

50

) of COX-2 by IC

50

concentrations of COX-1 [50, 51]. Valdecoxib possesses the highest COX-2 selectivity (0.009 [50]), followed by etodolac (0.021[50]), rofecoxib (0.029 [50]), meloxicam (0.032 [50]), celecoxib (0.036 [50]), etoricoxib (0.032 [50]), diclofenac (0.1 [50]), ibuprofen (2.89 [50]), naproxen (>10.87 [50]), and aspirin ( > 100 [51]). We categorized exposure to selective COX-2 inhibitors based on the number of pre- scriptions prior to the index date. We categorized subjects who did not receive any selective COX-2 drug prior to the index date as ‘ no prior use ’ (0 prescriptions, reference). Selective COX-2 users were classified into categories of 1–9, 10–29, or 30 prescriptions prior to the index date.

The number of prescriptions serves as an approximation of exposure duration, since an average prescription covers 45 to 90 days of treatment, depending on the number of tablets (1 or 2) taken per day. In a sensitivity analysis, we considered patients with < 5 prescriptions of selective COX- 2 inhibitors as non-users. We also investigated individual use of etodolac, meloxicam, celecoxib, rofecoxib, etoricoxib, valdecoxib, lumiracoxib and diclofenac employing the same exposure cate- gories as for the combined selective COX-2 inhibitor variable where possible.

Statistical analysis

We conducted conditional logistic regression analyses using the SAS statistical software version 9.4 (SAS Institute Inc, Cary, NC) to determine relative risks, estimated as odds ratios (ORs) with 95% confidence intervals (CIs), of glioma in relation to use of selective COX-2 inhibitors.

A two-sided p-value of <0.05 was considered statistically significant. We performed tests of linear trend by modelling the median value of each category as a continuous variable in the multivariate model, the coefficient for which was evaluated using a Wald test.

In univariate analyses, we investigated the influence of various potential confounding vari-

ables, including body mass index (BMI, <18.5 kg/m², 18.5–24.9 kg/m², 25.0–29.9 kg/m², 30.0

kg/m², unknown), smoking status (never, current, past, unknown) and presence versus absence

of specific medical conditions or diseases (congestive heart failure, diabetes, rheumatoid arthri-

tis, allergies (asthma and hay fever), and use of estrogens (0, 1 – 14, 15 prescriptions; women

only). We only included variables in the final multivariate analysis that altered the risk estimate

(4)

of glioma by > 10%. We adjusted our final analysis for BMI, smoking, diabetes, and congestive heart failure. We also included other NSAIDs (ibuprofen, naproxen, and aspirin) and non- NSAID analgesics (paracetamol, opioids) in the multivariate model. We stratified analyses by gender and presence of supporting READ codes (surgery, chemotherapy, radiotherapy, or referrals to a specialized oncology clinic). In further subanalyses, we restricted cases to glioblas- toma patients, used the combination of all NSAIDs as exposure variable, excluded diclofenac from the group of selective COX-2 inhibitors due to its lower affinity to COX-2, and investi- gated mutually exclusive use of selective COX-2 inhibitors (i.e., excluding patients using other NSAIDs from the study population).

Results

We identified 2,469 cases and 24,690 controls in the CPRD database. The mean age ± standard deviation (SD) at the time of glioma diagnosis was 55.3 ± 18.8 years. 44.7% of cases and con- trols were women. The mean number of years of active history in the database prior to the index date was 11.3 ± 5.2 years for cases and controls. Among patients with known WHO grade of glioma (58.1%), 1,079 patients (75.2%) were diagnosed as glioblastoma.

Table 1 displays characteristics of glioma cases and controls. As with prior data using the CPRD [49], underweight versus normal weight (OR = 0.42; 95% CI = 0.24–0.74), a history of congestive heart failure (OR = 0.58; 95% CI = 0.38 – 0.87) and a history of diabetes mellitus (OR = 0.81; 95% CI = 0.67–0.97) were related to reduced glioma risk. Current versus never smoking (OR = 0.87; 95% CI = 0.77 – 0.99) also showed an inverse relation to glioma. By com- parison, no associations with glioma were found for rheumatoid arthritis, history of allergies, or use of estrogens in women.

Table 2 provides ORs of glioma in association with use of selective COX-2 inhibitors or use of other NSAIDs or non-NSAID analgesics. There was no association between use of selective COX-2 inhibitors and risk of glioma. As compared with no prior use, the OR for 1–9 prescrip- tions was 1.02 (95% CI = 0.92 – 1.13), for 10 – 29 prescriptions it was 1.01 (95% CI = 0.80 – 1.28), and for 30 prescriptions it was 1.16 (95% CI = 0.86–1.55, p-value for trend = 0.408). Results were similar for aspirin and ibuprofen (Table 2).

We noted that as compared with no prior use, 10–29 prescriptions of naproxen or 1–9 pre- scriptions of paracetamol were inversely associated with glioma risk (OR = 0.52; 95%

CI = 0.28–0.96, OR = 0.80; 95% CI = 0.68–0.94, respectively), but there was no significant trend with increasing numbers of prescriptions (naproxen: p-value for trend = 0.692; paraceta- mol: p-value for trend = 0.403). Use of 1–9 prescriptions of opiates/opioids was positively asso- ciated with glioma risk (OR = 1.32; 95% CI = 1.12 – 1.56) but again, there was no detectable dose-response relation when investigating increasing duration of drug use (p-value for trend = 0.439) (Table 2). When analysing individual selective COX-2 inhibitors in relation to glioma risk, we found no significant associations with use of etodolac, meloxicam, celecoxib, rofecoxib, etoricoxib, valdecoxib, lumiracoxib, or diclofenac (Table 3). Frequencies of use of the specific COX-2 inhibiting drugs and the grouped variable of “selective COX-2 inhibitors”

vary due to successive or combined use of several selective COX-2 inhibitors.

There was no effect modification by sex or when restricting cases (and corresponding con-

trols) to glioblastoma patients or to glioma patients with READ codes for surgery, chemother-

apy, radiotherapy, or referrals to a specialized oncology clinic. The results remained similar to

those of our main analysis when we conducted analyses of patients with < 5 prescriptions of

selective COX-2 inhibitors as the non-exposed, when combining all NSAIDs, when excluding

diclofenac from the group of selective COX2 inhibitors, or when investigating mutually exclu-

sive use of selective COX-2 inhibitors (data not shown).

(5)

Discussion

To the best of our knowledge, the current analysis is the first observational study to report on use of selective COX-2 inhibitors specifically in relation to the risk of glioma. In contrast to a large number of laboratory studies showing potential preventive effects of selective COX-2 inhibitors on glioma pathogenesis, we found no association between selective COX-2 inhibitor use and glioma.

The relation of NSAID use to risk of glioma or brain cancer has been explored in a number of previous observational studies. Three case – control studies observed a statistically significant

Table 1. Characteristics of glioma cases and controls.

Variable Number of cases (%) (n= 2,469) Number of controls (%) (n= 24,690) OR (95% CI)* Age (years)*

0–9 72 (2.9) 728 (3.0) -

10–19 95 (3.9) 941 (3.8) -

20–29 113 (4.6) 1,122 (4.5) -

30–39 182 (7.4) 1,827 (7.4) -

40–49 273 (11.1) 2,765 (11.2) -

50–59 536 (21.7) 5,308 (21.5) -

60–69 608 (24.6) 6,151 (24.9) -

70–79 450 (18.2) 4,514 (18.3) -

80–90 140 (5.7) 1,334 (5.4) -

Sex*

Men 1,365 (55.3) 13,650 (55.3) -

Women 1,104 (44.7) 11,040 (44.7) -

BMI (kg/m²)

<18.5 13 (0.5) 289 (1.2) 0.42 (0.24–0.74)

18.5–24.9 753 (30.5) 7,119 (28.8) 1.00 (reference)

25–29.9 749 (30.3) 7,295 (29.6) 0.97 (0.87–1.08)

30.0 401 (16.2) 4,046 (16.4) 0.94 (0.83–1.07)

Unknown 553 (22.4) 5,941 (24.1) 0.84 (0.73–0.96)

Smoking status

Never smoker 1,185 (48.0) 11,071 (44.8) 1.00 (reference)

Current smoker 398 (16.1) 4,220 (17.1) 0.87 (0.77–0.99)

Past smoker 578 (23.4) 5,784 (23.4) 0.94 (0.84–1.05)

Unknown 308 (12.5) 3,615 (14.6) 0.69 (0.58–0.82)

Comorbidities

CHF 25 (1.0) 424 (1.7) 0.58 (0.38–0.87)

Diabetes 138 (5.6) 1,674 (6.8) 0.81 (0.67–0.97)

Rheumatoid arthritis 24 (1.0) 277 (1.1) 0.86 (0.57–1.32)

Allergies 89 (3.6) 944 (3.8) 0.94 (0.75–1.18)

Medication use Estrogens^#

No prior use 837 (75.8) 8,340 (75.5) 1.00 (reference)

1–14 Rx 177 (16.0) 1,812 (16.4) 1.03 (0.86–1.25)

15 Rx 90 (8.2) 888 (8.0) 1.04 (0.79–1.36)

*Matching variables: age, sex, general practice, and number of years of active history in the database.

#Women only. BMI: Body mass index; CHF: congestive heart failure; OR: Odds ratio, Rx: prescriptions.

doi:10.1371/journal.pone.0149293.t001

(6)

inverse association between use of NSAIDs and risk of glioma [40, 41] or glioblastoma [39]

Three other case-control studies [34, 38, 42], one prospective cohort study [37], one random- ized clinical trial [36], and one recent meta-analysis [35] found no association between NSAID use and brain tumour risk, and two previous cohort studies [43, 44] found an increased risk of brain tumours among NSAID users. Such heterogeneity of results from previous studies may

Table 2. Risk of glioma in patients using NSAIDs or other analgesics.

Variable Number of cases (%) (n= 2,469) Number of controls (%) (n= 24,690) Adjusted OR (95% CI)*^/#

Selective COX-2 inhibitors

No prior use 1,609 (65.2) 16,242 (65.8) 1.00 (reference)

1–9 Rx 710 (28.8) 6,982 (28.3) 1.02 (0.92–1.13)

10–29 Rx 91 (3.7) 939 (3.8) 1.01 (0.80–1.28)

30 Rx 59 (2.4) 527 (2.1) 1.16 (0.86–1.55)

p-value for trend 0.408

Aspirin

1–9 Rx 40 (1.6) 481 (2.0) 0.83 (0.60–1.15)

10–29 Rx 11 (0.5) 140 (0.6) 0.80 (0.43–1.50)

30 Rx 13 (0.5) 115 (0.5) 1.19 (0.66–2.13)

Ibuprofen

No prior use 1,676 (67,9) 16,607 (67,3) 1.00 (reference)

1–9 Rx 718 (29.1) 7,353 (29.8) 0.95 (0.86–1.05)

10–29 Rx 53 (2.2) 505 (2.1) 1.03 (0.77–1.39)

30 Rx 22 (0.9) 225 (0.9) 0.94 (0.60–1.48)

Naproxen

1–9 Rx 277 (11.2) 2,931 (11.9) 0.91 (0.79–1.05)

10–29 Rx 11 (0.5) 203 (0.8) 0.52 (0.28–0.96)

30 Rx 15 (0.6) 102 (0.4) 1.45 (0.83–2.52)

Non-NSAID analgesics Paracetamol

1–9 Rx 841 (34.1) 8,417 (34.1) 0.80 (0.68–0.94)

10–29 Rx 168 (6.8) 1,674 (6.8) 0.87 (0.66–1.14)

30 Rx 169 (6.8) 1,760 (7.1) 0.76 (0.55–1.06)

Opioids

1–9 Rx 853 (34.6) 8,106 (32.8) 1.32 (1.12–1.56)

10–29 Rx 135 (5.5) 1,405 (5,7) 1.19 (0.88–1.60)

30 Rx 162 (6.6) 1,623 (6.6) 1.35 (0.96–1.89)

*Matching variables: age, sex, general practice, and number of years of active history in the database.^#Adjustment for body mass index, smoking, diabetes, congestive heart failure, and all other medications in this table. COX: Cyclooxygenase; NSAIDs: non-steroidal anti-inﬂammatory drugs; Rx:

prescriptions.

(7)

be due to differences in study designs, exposure assessments (i.e., self-reported versus medical record data), underlying drug dosages, durations of use, and adjustments for covariates [35].

Table 3. Risk of glioma in patients using etodoloac, meloxicam, celecoxib, rofecoxib, etoricoxib, valdecoxib, lumiracoxib or diclofenac.

Variable Number of cases (%) (n= 2,469) Number of controls (%) (n= 24,690) Adjusted OR (95% CI)*^/#

Etodolac

1–9 Rx 21 (0.9) 180 (0.7) 0.79 (0.17–3.70)

10–29 Rx 5 (0.2) 20 (0.1) 1.77 (0.30–10.42)

30 Rx 2 (0.1) 13 (0.1) 0.64 (0.15–2.81)

Meloxicam

1–9 Rx 55 (2.2) 565 (2.3) 0.98 (0.73–1.31)

10–29 Rx 11 (0.5) 77 (0.3) 1.37 (0.71–2.66)

30 Rx 4 (0.2) 42 (0.2) 0.98 (0.34–2.82)

Celecoxib

1–9 Rx 60 (2.4) 613 (2.5) 1.00 (0.75–1.33)

10–29 Rx 10 (0.4) 71 (0.3) 1.51 (0.76–3.00)

30 Rx 4 (0.2) 23 (0.1) 1.67 (0.57–4.89)

Rofecoxib

1–9 Rx 52 (2.1) 530 (2.2) 0.99 (0.73–1.33)

10–29 Rx 9 (0.4) 88 (0.4) 0.92 (0.45–1.88)

30 Rx 1 (0.0) 21 (0.1) 0.61 (0.10–3.63)

Etoricoxib

1–9 Rx 26 (1.1) 223 (0.9) 1.12 (0.73–1.71)

10 Rx 2 (0.1) 43 (0.2) 0.50 (0.12–2.05)

Valdecoxib

1 Rx 2 (0.1) 27 (0.1) 0.80 (0.19–3.36)

Lumiracoxib

1 Rx 1 (0.0) 6 (0.0) 1.28 (0.10–16.42)

Diclofenac

No prior use 1,683 (68,2) 16,902 (68,5) 1.00 (reference)

1–9 Rx 676 (27.4) 6,733 (27.3) 1.00 (0.90–1.11)

10–29 Rx 70 (2.8) 690 (2.8) 1.05 (0.81–1.36)

30 Rx 40 (1,6) 365 (1.5) 1.09 (0.77–1.54)

*Matching variables: age, sex, general practice, and number of years of active history in the database.^#Adjustment for body mass index, smoking, diabetes, congestive heart failure, all other medications in this table andTable 2except the combined variable of selective COX-2 inhibitors.

(8)

None of the aforementioned studies [34 – 44] investigated the association between use of selective COX-2 inhibitors specifically and glioma risk. A large number of laboratory investiga- tions reported on inhibitory effects of selective COX-2 inhibitors on various stages of glioma pathogenesis [3–6, 8–13, 15, 18–23], prompting clinical trials to investigate selective COX-2 inhibitors as adjuvant therapy for glioma [24 – 32]. The discrepancy between the laboratory findings and our results may be explained by a number of factors. First, despite mechanistic data regarding protective effects of selective COX-2 inhibitors on glioma development [33], most preclinical studies were designed to investigate therapeutic rather than preventive drug effects. In addition, drug doses used in in vitro experiments often exceeded the usual drug con- centrations reached in the brain, particularly in experiments involving intratumoral applica- tion of selective COX-2 inhibitors or coupling of selective COX-2 inhibitors to nanoparticles aimed at increasing drug concentrations in the tumour [52]. Phase I/II clinical trials were also unable to show a clear benefit of adjuvant therapy with selective COX-2 inhibitors [24, 25, 28 – 31], even though they were modelled on biological tumour models and used combined thera- peutic approaches.

Our null result for selective COX-2 inhibitors and glioma are in line with previous null results for glioma and use of diclofenac, which has a 4-fold higher selectivity for COX-2 than COX-1 [51].

We observed an increased glioma risk in association with short-term use of opiates/opioids.

Possible underlying biologic mechanisms remain speculative but may involve opiate-mediated activation of Mitogen-Activated-Kinases [53] or Akt kinases [54]. However, confounding by indication due to opiate or selective COX-2 inhibitors use in patients with pre-existing but undiagnosed glioma or glioblastoma is another possible explanation, despite the fact that we shifted the index date backward in time by one year for cases and controls.

Certain shortcomings of our study warrant discussion. While patients using selective COX- 2 inhibitors require drug prescriptions, other NSAIDs such as aspirin, ibuprofen, and naproxen are available over the counter (OTC) in the UK [55]. We therefore may have underestimated NSAID use in patients taking OTC medications, which could have biased the risk estimates towards the null value. However, a prior study showed that prescription-based data adequately reflect use of OTC medications [56]. Also, we had rather small sample sizes in some exposure categories, such as long-term intake of selective COX-2 inhibitors among cases. We were unable to closely control for socioeconomic status, and we were not able to take physical activ- ity or other lifestyle factors possibly related to glioma into account in our analyses, because these variables are not regularly recorded in the CPRD. However, prior studies have not revealed meaningful associations between adulthood lifestyle factors and glioma risk [57, 58].

Our findings may not apply to non-Caucasian populations, because 86% of individuals in our database are Caucasian [59].

Our study has a number of important strengths. The CPRD is a well-established, large-, and validated database [46]. Because we generated cases and controls from a pre-existing database, selection bias is unlikely. Further, recall bias is absent because the data regarding medications and concomitant diseases were collected prospectively. Finally, only patients with an active his- tory of at least three years in the CPRD database were included in the current study in order to increase the likelihood of capturing only newly diagnosed cases and to ensure a sufficiently long history of exposure to selective COX-2 inhibitor use.

In summary, we found no association between use of selective COX-2 inhibitors and the

risk of glioma. Further observational studies are warranted to confirm our findings and to help

explain the apparent discrepancies between laboratory and observational data.

(9)

Acknowledgments

We thank Pascal Egger for technical support and programming and all members of the KFO 262 for critical discussions regarding our manuscript.

Author Contributions

Conceived and designed the experiments: CS CM CB SJ PH ML. Performed the experiments:

CS. Analyzed the data: CS. Contributed reagents/materials/analysis tools: CM SJ ML. Wrote the paper: CS CM CB SJ UB PH ML.

References

1. Stupp R, Hegi ME, Mason WP, van den Bent MJ, Taphoorn MJ, Janzer RC, et al. Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial. The lancet oncology. 2009 May; 10(5):459–66. doi:10.1016/S1470-2045(09)70025-7PMID:19269895

2. Inskip PD, Linet MS, Heineman EF. Etiology of brain tumors in adults. Epidemiologic reviews. 1995; 17 (2):382–414. PMID:8654518

3. Sareddy GR, Geeviman K, Ramulu C, Babu PP. The nonsteroidal anti-inflammatory drug celecoxib suppresses the growth and induces apoptosis of human glioblastoma cells via the NF-kappaB path- way. Journal of neuro-oncology. 2012 Jan; 106(1):99–109. doi:10.1007/s11060-011-0662-xPMID:

21847707

4. Sharma V, Dixit D, Ghosh S, Sen E. COX-2 regulates the proliferation of glioma stem like cells. Neuro- chemistry international. 2011 Oct; 59(5):567–71. doi:10.1016/j.neuint.2011.06.018PMID:21763744 5. Gaiser T, Becker MR, Habel A, Reuss DE, Ehemann V, Rami A, et al. TRAIL-mediated apoptosis in

malignant glioma cells is augmented by celecoxib through proteasomal degradation of survivin. Neuro- science letters. 2008 Sep 12; 442(2):109–13. doi:10.1016/j.neulet.2008.07.014PMID:18634847 6. Chen JC, Chen Y, Su YH, Tseng SH. Celecoxib increased expression of 14-3-3sigma and induced

apoptosis of glioma cells. Anticancer research. 2007 Jul-Aug; 27(4B):2547–54. PMID:17695552 7. Kang SG, Kim JS, Park K, Kim JS, Groves MD, Nam DH. Combination celecoxib and temozolomide in

C6 rat glioma orthotopic model. Oncology reports. 2006 Jan; 15(1):7–13. PMID:16328028

8. Kardosh A, Blumenthal M, Wang WJ, Chen TC, Schonthal AH. Differential effects of selective COX-2 inhibitors on cell cycle regulation and proliferation of glioblastoma cell lines. Cancer biology & therapy.

2004 Jan; 3(1):55–62.

9. Nam DH, Park K, Park C, Im YH, Kim MH, Lee S, et al. Intracranial inhibition of glioma cell growth by cyclooxygenase-2 inhibitor celecoxib. Oncology reports. 2004 Feb; 11(2):263–8. PMID:14719052 10. Suzuki K, Gerelchuluun A, Hong Z, Sun L, Zenkoh J, Moritake T, et al. Celecoxib enhances radiosensi-

tivity of hypoxic glioblastoma cells through endoplasmic reticulum stress. Neuro-oncology. 2013 Sep;

15(9):1186–99. doi:10.1093/neuonc/not062PMID:23658321

11. Ma HI, Chiou SH, Hueng DY, Tai LK, Huang PI, Kao CL, et al. Celecoxib and radioresistant glioblastoma-derived CD133+ cells: improvement in radiotherapeutic effects. Laboratory investigation. Journal of neurosurgery. 2013 Mar; 114(3):651–62.

12. Kuipers GK, Slotman BJ, Wedekind LE, Stoter TR, Berg J, Sminia P, et al. Radiosensitization of human glioma cells by cyclooxygenase-2 (COX-2) inhibition: independent on COX-2 expression and depen- dent on the COX-2 inhibitor and sequence of administration. International journal of radiation biology.

2007 Oct; 83(10):677–85. PMID:17729162

13. Bijnsdorp IV, van den Berg J, Kuipers GK, Wedekind LE, Slotman BJ, van Rijn J, et al. Radiosensitizing potential of the selective cyclooygenase-2 (COX-2) inhibitor meloxicam on human glioma cells. Journal of neuro-oncology. 2007 Oct; 85(1):25–31. PMID:17447009

14. Kang KB, Wang TT, Woon CT, Cheah ES, Moore XL, Zhu C, et al. Enhancement of glioblastoma radio- response by a selective COX-2 inhibitor celecoxib: inhibition of tumor angiogenesis with extensive tumor necrosis. International journal of radiation oncology, biology, physics. 2007 Mar 1; 67(3):888–96.

PMID:17293239

15. Petersen C, Petersen S, Milas L, Lang FF, Tofilon PJ. Enhancement of intrinsic tumor cell radiosensitiv- ity induced by a selective cyclooxygenase-2 inhibitor. Clin Cancer Res. 2000 Jun; 6(6):2513–20. PMID:

10873107

16. Kosaka A, Ohkuri T, Okada H. Combination of an agonistic anti-CD40 monoclonal antibody and the COX-2 inhibitor celecoxib induces anti-glioma effects by promotion of type-1 immunity in myeloid cells

(10)

and T-cells. Cancer Immunol Immunother. 2014 Aug; 63(8):847–57. doi:10.1007/s00262-014-1561-8 PMID:24878890

17. Robison NJ, Campigotto F, Chi SN, Manley PE, Turner CD, Zimmerman MA, et al. A phase II trial of a multi-agent oral antiangiogenic (metronomic) regimen in children with recurrent or progressive cancer.

Pediatric blood & cancer. 2014 Apr; 61(4):636–42.

18. Wagemakers M, van der Wal GE, Cuberes R, Alvarez I, Andres EM, Buxens J, et al. COX-2 Inhibition Combined with Radiation Reduces Orthotopic Glioma Outgrowth by Targeting the Tumor Vasculature.

Translational oncology. 2009 Mar; 2(1):1–7. PMID:19252746

19. Tuettenberg J, Grobholz R, Korn T, Wenz F, Erber R, Vajkoczy P. Continuous low-dose chemotherapy plus inhibition of cyclooxygenase-2 as an antiangiogenic therapy of glioblastoma multiforme. Journal of cancer research and clinical oncology. 2005 Jan; 131(1):31–40. PMID:15565458

20. Eberstal S, Fritzell S, Sanden E, Visse E, Darabi A, Siesjo P. Immunizations with unmodified tumor cells and simultaneous COX-2 inhibition eradicate malignant rat brain tumors and induce a long-lasting CD8(+) T cell memory. Journal of neuroimmunology. 2014 Sep 15; 274(1–2):161–7. doi:10.1016/j.

jneuroim.2014.06.019PMID:25022336

21. Zhang H, Tian M, Xiu C, Wang Y, Tang G. Enhancement of antitumor activity by combination of tumor lysate-pulsed dendritic cells and celecoxib in a rat glioma model. Oncology research. 2013; 20 (10):447–55. PMID:24308155

22. Eberstal S, Sanden E, Fritzell S, Darabi A, Visse E, Siesjo P. Intratumoral COX-2 inhibition enhances GM-CSF immunotherapy against established mouse GL261 brain tumors. International journal of cancer. 2014 Jun 1; 134(11):2748–53.

23. Eberstal S, Badn W, Fritzell S, Esbjornsson M, Darabi A, Visse E, et al. Inhibition of cyclooxygenase-2 enhances immunotherapy against experimental brain tumors. Cancer Immunol Immunother. 2012 Aug;

61(8):1191–9. doi:10.1007/s00262-011-1196-yPMID:22213142

24. Andre N, Abed S, Orbach D, Alla CA, Padovani L, Pasquier E, et al. Pilot study of a pediatric metronomic 4-drug regimen. Oncotarget. 2011 Dec; 2(12):960–5. PMID:22156656

25. Gilbert MR, Gonzalez J, Hunter K, Hess K, Giglio P, Chang E, et al. A phase I factorial design study of dose-dense temozolomide alone and in combination with thalidomide, isotretinoin, and/or celecoxib as postchemoradiation adjuvant therapy for newly diagnosed glioblastoma. Neuro-oncology. 2010 Nov;

12(11):1167–72. doi:10.1093/neuonc/noq100PMID:20729242

26. Walbert T, Gilbert MR, Groves MD, Puduvalli VK, Yung WK, Conrad CA, et al. Combination of 6-thio- guanine, capecitabine, and celecoxib with temozolomide or lomustine for recurrent high-grade glioma.

Journal of neuro-oncology. 2011 Apr; 102(2):273–80. doi:10.1007/s11060-010-0313-7PMID:

20652724

27. Stockhammer F, Misch M, Koch A, Czabanka M, Plotkin M, Blechschmidt C, et al. Continuous low- dose temozolomide and celecoxib in recurrent glioblastoma. Journal of neuro-oncology. 2010 Dec; 100 (3):407–15. doi:10.1007/s11060-010-0192-yPMID:20446016

28. Kesari S, Schiff D, Henson JW, Muzikansky A, Gigas DC, Doherty L, et al. Phase II study of temozolomide, thalidomide, and celecoxib for newly diagnosed glioblastoma in adults. Neuro-oncology. 2008 Jun; 10(3):300–8. doi:10.1215/15228517-2008-005PMID:18403492

29. Hau P, Kunz-Schughart L, Bogdahn U, Baumgart U, Hirschmann B, Weimann E, et al. Low-dose chemotherapy in combination with COX-2 inhibitors and PPAR-gamma agonists in recurrent high-grade gliomas—a phase II study. Oncology. 2007; 73(1–2):21–5. doi:10.1159/000120028PMID:18332649 30. Kesari S, Schiff D, Doherty L, Gigas DC, Batchelor TT, Muzikansky A, et al. Phase II study of metro-

nomic chemotherapy for recurrent malignant gliomas in adults. Neuro-oncology. 2007 Jul; 9(3):354–63.

PMID:17452651

31. Levin VA, Giglio P, Puduvalli VK, Jochec J, Groves MD, Yung WK, et al. Combination chemotherapy with 13-cis-retinoic acid and celecoxib in the treatment of glioblastoma multiforme. Journal of neuro- oncology. 2006 May; 78(1):85–90. PMID:16391896

32. Reardon DA, Quinn JA, Vredenburgh J, Rich JN, Gururangan S, Badruddoja M, et al. Phase II trial of iri- notecan plus celecoxib in adults with recurrent malignant glioma. Cancer. 2005 Jan 15; 103(2):329–38.

PMID:15558802

33. Fujita M, Kohanbash G, Fellows-Mayle W, Hamilton RL, Komohara Y, Decker SA, et al. COX-2 block- ade suppresses gliomagenesis by inhibiting myeloid-derived suppressor cells. Cancer research. 2011 Apr 1; 71(7):2664–74. doi:10.1158/0008-5472.CAN-10-3055PMID:21324923

34. Bannon FJ, O'Rorke MA, Murray LJ, Hughes CM, Gavin AT, Fleming SJ, et al. Non-steroidal anti- inflammatory drug use and brain tumour risk: a case-control study within the Clinical Practice Research Datalink. Cancer Causes Control. 2013 Nov; 24(11):2027–34. doi:10.1007/s10552-013-0279-9PMID:

23990380

(11)

35. Liu Y, Lu Y, Wang J, Xie L, Li T, He Y, et al. Association between nonsteroidal anti-inflammatory drug use and brain tumour risk: a meta-analysis. British journal of clinical pharmacology. 2014 Jul; 78(1):58– 68. doi:10.1111/bcp.12311PMID:24341448

36. Cook NR, Lee IM, Gaziano JM, Gordon D, Ridker PM, Manson JE, et al. Low-dose aspirin in the primary prevention of cancer: the Women's Health Study: a randomized controlled trial. Jama. 2005 Jul 6;

294(1):47–55. PMID:15998890

37. Daugherty SE, Moore SC, Pfeiffer RM, Inskip PD, Park Y, Hollenbeck A, et al. Nonsteroidal anti-inflammatory drugs and glioma in the NIH-AARP Diet and Health Study cohort. Cancer prevention research (Philadelphia, Pa. 2011 Dec; 4(12):2027–34.

38. Gaist D, Garcia-Rodriguez LA, Sorensen HT, Hallas J, Friis S. Use of low-dose aspirin and non-aspirin nonsteroidal anti-inflammatory drugs and risk of glioma: a case-control study. British journal of cancer.

2013 Mar 19; 108(5):1189–94. doi:10.1038/bjc.2013.87PMID:23449355

39. Sivak-Sears NR, Schwartzbaum JA, Miike R, Moghadassi M, Wrensch M. Case-control study of use of nonsteroidal antiinflammatory drugs and glioblastoma multiforme. American journal of epidemiology.

2004 Jun 15; 159(12):1131–9. PMID:15191930

40. Ferris JS, McCoy L, Neugut AI, Wrensch M, Lai R. HMG CoA reductase inhibitors, NSAIDs and risk of glioma. International journal of cancer. 2012 Sep 15; 131(6):E1031–7.

41. Scheurer ME, El-Zein R, Thompson PA, Aldape KD, Levin VA, Gilbert MR, et al. Long-term anti-inflammatory and antihistamine medication use and adult glioma risk. Cancer Epidemiol Biomarkers Prev.

2008 May; 17(5):1277–81. doi:10.1158/1055-9965.EPI-07-2621PMID:18483351

42. Scheurer ME, Amirian ES, Davlin SL, Rice T, Wrensch M, Bondy ML. Effects of antihistamine and anti- inflammatory medication use on risk of specific glioma histologies. International journal of cancer. 2011 Nov 1; 129(9):2290–6.

43. Sorensen HT, Friis S, Norgard B, Mellemkjaer L, Blot WJ, McLaughlin JK, et al. Risk of cancer in a large cohort of nonaspirin NSAID users: a population-based study. British journal of cancer. 2003 Jun 2; 88(11):1687–92. PMID:12771981

44. Friis S, Sorensen HT, McLaughlin JK, Johnsen SP, Blot WJ, Olsen JH. A population-based cohort study of the risk of colorectal and other cancers among users of low-dose aspirin. British journal of cancer. 2003 Mar 10; 88(5):684–8. PMID:12618874

45. Walley T, Mantgani A. The UK General Practice Research Database. Lancet. 1997 Oct 11; 350 (9084):1097–9. PMID:10213569

46. Jick SS, Kaye JA, Vasilakis-Scaramozza C, Garcia Rodriguez LA, Ruigomez A, Meier CR, et al. Valid- ity of the general practice research database. Pharmacotherapy. 2003 May; 23(5):686–9. PMID:

12741446

47. Khan NF, Harrison SE, Rose PW. Validity of diagnostic coding within the General Practice Research Database: a systematic review. Br J Gen Pract. 2010 Mar; 60(572):e128–36. doi:10.3399/

bjgp10X483562PMID:20202356

48. Louis DN, Ohgaki H, Wiestler OD, Cavenee WK, Burger PC, Jouvet A, et al. The 2007 WHO classification of tumours of the central nervous system. Acta neuropathologica. 2007 Aug; 114(2):97–109.

PMID:17618441

49. Seliger C, Ricci C, Meier CR, Bodmer M, Jick SS, Bogdahn U, et al. Diabetes, use of antidiabetic drugs, and the risk of glioma. Neuro-oncology. 2015 Jun 20.

50. Gierse JK, Zhang Y, Hood WF, Walker MC, Trigg JS, Maziasz TJ, et al. Valdecoxib: assessment of cyclooxygenase-2 potency and selectivity. The Journal of pharmacology and experimental therapeu- tics. 2005 Mar; 312(3):1206–12. PMID:15494548

51. Warner TD, Giuliano F, Vojnovic I, Bukasa A, Mitchell JA, Vane JR. Nonsteroid drug selectivities for cyclo-oxygenase-1 rather than cyclo-oxygenase-2 are associated with human gastrointestinal toxicity:

a full in vitro analysis. Proceedings of the National Academy of Sciences of the United States of Amer- ica. 1999 Jun 22; 96(13):7563–8. PMID:10377455

52. Vera M, Barcia E, Negro S, Marcianes P, Garcia-Garcia L, Slowing K, et al. New celecoxib multiparticu- late systems to improve glioblastoma treatment. International journal of pharmaceutics. 2014 Oct 1;

473(1–2):518–27. doi:10.1016/j.ijpharm.2014.07.028PMID:25066075

53. Gutstein HB, Rubie EA, Mansour A, Akil H, Woodgett JR. Opioid effects on mitogen-activated protein kinase signaling cascades. Anesthesiology. 1997 Nov; 87(5):1118–26. PMID:9366464

54. Heiss A, Ammer H, Eisinger DA. delta-Opioid receptor-stimulated Akt signaling in neuroblastoma x glioma (NG108-15) hybrid cells involves receptor tyrosine kinase-mediated PI3K activation. Experimental cell research. 2009 Jul 15; 315(12):2115–25. doi:10.1016/j.yexcr.2009.04.002PMID:19362548 55. http://www.nhs.uk.

(12)

56. Yood MU, Campbell UB, Rothman KJ, Jick SS, Lang J, Wells KE, et al. Using prescription claims data for drugs available over-the-counter (OTC). Pharmacoepidemiology and drug safety. 2007 Sep; 16 (9):961–8. PMID:17654746

57. Moore SC, Rajaraman P, Dubrow R, Darefsky AS, Koebnick C, Hollenbeck A, et al. Height, body mass index, and physical activity in relation to glioma risk. Cancer research. 2009 Nov 1; 69(21):8349–55.

doi:10.1158/0008-5472.CAN-09-1669PMID:19808953

58. Benson VS, Pirie K, Green J, Casabonne D, Beral V. Lifestyle factors and primary glioma and meningi- oma tumours in the Million Women Study cohort. British journal of cancer. 2008 Jul 8; 99(1):185–90.

doi:10.1038/sj.bjc.6604445PMID:18560401

59. 2011 UKC. United Kingdom population by ethnic group. UK: Office for national Statistics Newport.